Accurate regulatory reporting has been an ongoing and ever-increasing challenge for financial institutions causing incredible financial toll, security challenges and strain on banks and their employees. However, the introduction of new technology, namely artificial intelligence (AI) could ease these burdens and allow companies to dedicate resources elsewhere.
Sarva Srinivasan is the co-founder and President of EZOPS, an AI-powered data control and automation platform driving transformative operational efficiency for financial firms. With a strong understanding of artificial intelligence and machine learning, Srinivasan explained to The Fintech Times how this technology could transform regulatory reporting in banks:
Challenges
US banks alone have been fined a staggering $243billion since 2008, according to a Deloitte report. A Risk Management Association Survey estimates that 50 per cent of respondents spent six to ten per cent of their revenue on compliance costs. According to a report by Rice University’s Baker Institute for Public Policy the total cost of non-interest expenses, such as hiring new compliance officers or bringing in outside lawyers or consultants, increased after 2010 by an estimated $64.5billion per year. The burden of non-compliance is not only cost prohibitive but also stifles employee creativity. By constraining their ability to design products and improve business services, the regulatory burden can lead to burnout and lower employee retention. Business is further affected as the costs are inevitably passed onto customers who could choose to seek financial services elsewhere.
Another increasing challenge for financial institutions is the rapidly evolving regulatory standards. The number of regulations applied to US banks has more than doubled in the aftermath of the 2008 financial crisis with the passage of the Dodd-Frank Act of 2010. Concurrently, business volumes have exploded making the process of reporting more complicated and onerous.
Potential approaches to address these challenges
Traditionally banks and financial institutions have built rule-based software solutions to help them manage the quality of data as it flows through the data reporting lifecycle. Unfortunately, they are hard to update, maintain and deploy with changing business and regulatory conditions, resulting in ever-increasing costs and delays.
To counter some of these challenges, banks can deploy AI (Artificial Intelligence) enabled solutions and ML (Machine Learning) models to support regulatory processes and operations. AI has been leveraged in a number of operational processes within banks including fraud management, services monitoring etc., to boost efficiency, remove user error and help users to stay nimble and proactive. Similarly, Supervised and Unsupervised AI models can be leveraged to provide classification and pattern recognition capabilities to streamline regulatory reporting.
Anomaly Detection
Efficiency of a process is largely determined by the quality data used therein. Detecting anomalies as close to the data source as feasible will ensure that data issues are captured prior to transmission to downstream systems. AI models like autoencoder (a type of neural network) have been in existence for a few decades and can identify patterns of data as they flow through the models. When the model detects a dataset that does not fit a pattern, it tags the data and alerts users and other downstream systems. Without an AI model the only option is for users to individually define and code the many rules needed to identify the anomalies. These rules can be very complex, involve a number of data sources and often vary by product type, jurisdiction and counterparty. Given the speed of change in rules and types of products traded, institutions end up forever playing catchup. A similar approach can be adopted before data is sent to the regulators, as data quality issues could have crept in as part of the data transformations and rules being applied to the source data or during the reporting process and manual interventions. AI models look for patterns in data reported over a period of time and ensure that data inconsistencies are addressed intra-day as opposed to T+1 or T+2 as is generally the case.
Predict Break Reasons
Most banks are mandated to perform independent validation of data they have reported. This requires carrying out reconciliations between the data reported to regulators and the data in their reporting software solutions. Breaks (exceptions) identified as part of the process accumulate over a period of time and run into hundreds of thousands, whose resolution is largely a manually intensive process. Incorporating frameworks like XGBoost to provide supervised learning and classification features can tag breaks once reconciliation is completed. This can help users save time and effort in triaging and identifying reasons for breaks.
These approaches have shown to improve operational efficiency by over 35 per cent. By incorporating AI models that can operate on industry-wide data, from multiple banks and financial institutions, regulators can keep tabs on the pulse of the overall financial environment more effectively and reduce industry-wide risk.
AI and Automation
AI and Automation are becoming two sides of the same coin. By incorporating AI elements as part of process automation – in reporting, breaks resolution and client interactions, banks can preempt issues, improve operational efficiency and proactively address process and training needs for their teams
Summary
In an ever-changing landscape of regulatory reporting, banks and financial institutions have to stay ahead of the game to improve their accuracy of reporting and efficiency of operations in the face of business competition.
AI-enabled software that are low code, intuitive to configure and easy to manage gives businesses better control over the quality of data and related business processes.
